seamls2019 icon indicating copy to clipboard operation
seamls2019 copied to clipboard

Materials from the South East Asia Machine Learning School 2019 event

Introduction

This repository contains slides presented at the South East Asia Machine Learning School (SEA MLS) 2019. SEA MLS 2019 was a 5-day event held at Universitas Indonesia, Depok, Greater Jakarta, Indonesia between July 8 to July 12, 2019. It had the following sessions:

  • Lecture sessions from top machine learning experts and invited speakers (both basic and more advanced topics were covered);
  • Interactive practical sessions where participants had the chance to directly try out and implement models described in the lectures;
  • Panel sessions with machine learning researchers, practitioners, industry leaders, and government representatives about current challenges in AI;
  • Research presentation sessions where participants showcased their past work and projects to get feedback from fellow participants and invited speakers; and
  • Social events with the event's sponsors.

Content

The slides are organized according to the day they were presented as follows:

  • Day 1
    • Math foundations (expectation, differentiation, linear algebra, etc.)
    • Introduction to machine learning (supervised, unsupervised learning, applications)
    • Machine learning basics (overfitting, model selection, adding/reducing features, regularisation)
    • Practical I: Tensorflow and Colab
  • Day 2
    • Simple unsupervised learning (k-Means and Gaussian mixture models, matrix factorization, LDA, etc.)
    • Neural network basics (feedforward networks)
    • Tricks of the trade: optimisers, dropout, batch norm, layer norm
  • Day 3
    • Language models, word embeddings. Sequential predictions, RNNs, and LSTMs.
    • Sequence-to-sequence models. Conditional language generation (+attention)
    • Practical II: NLP in Tensorflow
  • Day 4
    • Convolutional neural networks (CNNs)
    • Multimodal machine learning: Grounding, VQA, etc.
    • Practical III: CNNs
  • Day 5
    • Deep probabilistic graphical models
    • Attention, self-attention, transformers, BERT
    • Deep learning for biomedicine